RS27 - Aortic Strain and Diameter After Endovascular and Open Thoracic Aortic Repair

Objectives: Although most well-known for its conduit function, the aorta has a crucial capacitance function for maintaining diastolic pressure and systemic perfusion throughout the cardiac cycle. Open and thoracic endovascular aortic repair (TEVAR) affects this capacitance function, but quantifying and describing these physiologic changes and potential downstream clinical implications has been limited to animal models or case reports and small cohort studies in humans. We developed a deep learning approach using echocardiography to quantify the aortic capacitance in the largest post-surgical human cohort to date. Our model and methods address the major limitations of quantification, reproducibility and clinical volume.

Methods: We used 51,730 echocardiogram videos from a quaternary referral center, divided into training, validation, and test cohorts, to train an AI model for frame-level measurement of aortic root diameter. Strain was defined as the change in root diameter divided by the minimum diameter during each cardiac cycle. TEVAR, open repair (OR), and age- and sex-matched control groups were assessed using this AI measurement approach. The Kruskal-Wallis test was used to compare patient-level data and the Dunn-Bonferroni test was used to identify differences between groups in aortic strain and diameter.

Results: The model accurately measured aortic root diameter with a lower mean absolute error (MAE) than intra-observer variation in measurement (MAE of 2.5mm vs. 2.9mm, p=0.010). A total of control (n=756) , TEVAR (n=216) and OR (n=2986) echocardiogram videos were evaluated from control (n=455), TEVAR (n=56) and OR (n=732) patients. Post-TEVAR patients had significantly lower aortic strain (median: 0.032, IQR: 0.022, p = 0.030) and larger aortic diameters (median: 3.564cm, IQR: 0.555cm, p < 0.001) compared to control strains (median: 0.042, IQR: 0.030) and diameters (median: 3.298cm, IQR: 0.494cm). (Figure 1). The direction of the diameter relationship was also observed in patients who underwent open thoracic aortic repair, although the magnitude was less (p < 0.001).

Conclusions: Deep learning quantification of aortic root diameter and aortic strain produces reproducible and precise measurements that can distinguish between patients who undergo TEVAR or OR vs. controls. Post-procedural patients had reduced aortic strain and larger aortic root diameters compared to age and sex matched controls, consistent with a loss of elasticity at the root observed in animal models of aortic dysfunction and human natural history studies of aortic aging. Further human validation and translational biomechanical studies are needed to interrogate these observations after aortic repair.